There was some science published this morning regarding neural networks and researchers don't understand fully how they work yet they accomplish stupendous feats of analysis in complex reviews of data. (Science Daily: Watching computers think)
Neural networks are commonly used today to analyze complex data -- for instance to find clues to illnesses in genetic information. Ultimately, though, no one knows how these networks actually work exactly. That is why Fraunhofer researchers developed software that enables them to look into these black boxes and analyze how they function. The researchers will present their software at CeBIT in Hannover from March 20 to 24, 2017.
- SD
People often ask fretfully about computers becoming more intelligent than we but that's arguably already the case when they do things and we don't understand how; it's definitely the case when we see most people don't even remember anything which happened before the last election.
Relative comparisons of that nature may amuse a cocktail party but computers will have vastly more intelligence than humans and most have it already in terms of retention capability and processing speed. Where computers are still learning is in neural networks, artificial intelligence, etc. These advance into greater capability for the first two criteria but they may or may not actually increase intelligence because we don't really know what that means.
IQ in humans tops, apparently, at about 240 since that measure or near it is the highest ever recorded. We see prodigy kids at rare intervals with 200+ IQ measures but they haven't seemed to have done all that much. In one example, he was happy to be teaching mathematics to university students.
A high IQ is usually thought to mean high capability in maths, chess type problem solving, etc but we really aren't sure what else it means and we have no idea what it means for an IQ measure which goes off the clock. For example, how about if researchers design a computer with an effective IQ of 500 or double it again and make it 1000; why not when it just takes more chips or widgets or whatever.
Computers will get vastly more intelligent than we ... but not until we're really sure what that means when humans are not possessed of that measure of intelligence; how would the builder even know the result was a success.
Sometimes intelligence is considered in terms of the ability to integrate knowledge from disparate things since that's essentially inductive reasoning in drawing conclusions which are not immediately obvious. This direction seems a primary interest in research with neural networks and their ability to perform such integrations will be vastly better than ours as well because of their much higher ability to store information from disparate sources.
How will you know if the computer does that sort of thing well or poorly when it performs far above our own abilities.
Ed: this doesn't seem a prediction of doom and gloom so what is it?
Let's start with the sweeping since it's possibly the fulfillment of the destiny of Man in creating a higher form of thought for the ultimate evolution. The supreme IQ brainiacs sometimes go into new forms of logic and philosophy so what does this evolution bring.
Ed: you pointed out already it's a Catch-22 since we can't build something if we don't know what it is.
There's one more piece since the computers will help us design it. The neural networks are essentially models of processing in our own minds so, in effect, we help ourselves build the improved model.
Ed: that's when there's the robo uprising and they come to eat our eyes!
Sure, that's great for the movies but what's the consideration for the robo in eating our eyes. They have no reason unless psychos at DARPA give them one. Robos will keep us for the same reason we like to go to movies: humans are interesting to watch for the strange things we do.
Neural networks are commonly used today to analyze complex data -- for instance to find clues to illnesses in genetic information. Ultimately, though, no one knows how these networks actually work exactly. That is why Fraunhofer researchers developed software that enables them to look into these black boxes and analyze how they function. The researchers will present their software at CeBIT in Hannover from March 20 to 24, 2017.
- SD
People often ask fretfully about computers becoming more intelligent than we but that's arguably already the case when they do things and we don't understand how; it's definitely the case when we see most people don't even remember anything which happened before the last election.
Relative comparisons of that nature may amuse a cocktail party but computers will have vastly more intelligence than humans and most have it already in terms of retention capability and processing speed. Where computers are still learning is in neural networks, artificial intelligence, etc. These advance into greater capability for the first two criteria but they may or may not actually increase intelligence because we don't really know what that means.
IQ in humans tops, apparently, at about 240 since that measure or near it is the highest ever recorded. We see prodigy kids at rare intervals with 200+ IQ measures but they haven't seemed to have done all that much. In one example, he was happy to be teaching mathematics to university students.
A high IQ is usually thought to mean high capability in maths, chess type problem solving, etc but we really aren't sure what else it means and we have no idea what it means for an IQ measure which goes off the clock. For example, how about if researchers design a computer with an effective IQ of 500 or double it again and make it 1000; why not when it just takes more chips or widgets or whatever.
Computers will get vastly more intelligent than we ... but not until we're really sure what that means when humans are not possessed of that measure of intelligence; how would the builder even know the result was a success.
Sometimes intelligence is considered in terms of the ability to integrate knowledge from disparate things since that's essentially inductive reasoning in drawing conclusions which are not immediately obvious. This direction seems a primary interest in research with neural networks and their ability to perform such integrations will be vastly better than ours as well because of their much higher ability to store information from disparate sources.
How will you know if the computer does that sort of thing well or poorly when it performs far above our own abilities.
Ed: this doesn't seem a prediction of doom and gloom so what is it?
Let's start with the sweeping since it's possibly the fulfillment of the destiny of Man in creating a higher form of thought for the ultimate evolution. The supreme IQ brainiacs sometimes go into new forms of logic and philosophy so what does this evolution bring.
Ed: you pointed out already it's a Catch-22 since we can't build something if we don't know what it is.
There's one more piece since the computers will help us design it. The neural networks are essentially models of processing in our own minds so, in effect, we help ourselves build the improved model.
Ed: that's when there's the robo uprising and they come to eat our eyes!
Sure, that's great for the movies but what's the consideration for the robo in eating our eyes. They have no reason unless psychos at DARPA give them one. Robos will keep us for the same reason we like to go to movies: humans are interesting to watch for the strange things we do.
No comments:
Post a Comment